Re: [PATCH 1/3] vfio-user: support VFIO_USER_DEVICE_FEATURE

[John Levon <john.levon@nutanix.com>]

On Thu, Apr 09, 2026 at 01:50:58PM +0200, Cédric Le Goater wrote:

> > +        /*
> > +         * Client expects ENOTTY for "not supported", but the protocol may
> > +         * return EINVAL (which should only occur in the case the feature isn't
> > +         * actually supported on the server).
> > +         */
> > +        if (msgp->hdr.error_reply == EINVAL) {
> > +            return -ENOTTY;
> > +        }
> 
> May be this is not necessary anymore with :
> 
>   https://urldefense.proofpoint.com/v2/url?u=https-3A__lore.kernel.org_qemu-2Ddevel_20260409114312.1704062-2D1-2Dclg-40redhat.com_&d=DwICaQ&c=s883GpUCOChKOHiocYtGcg&r=v7SNLJqx7b9Vfc7ZO82Wg4nnZ8O5XkACFQ30bVKxotI&m=lpMrwUH_bH0I_KS0ucNhIf9acrnTdwuaRPGfJXRAj2yImpt5jpWGpP82gP4MJmRZ&s=9mI6zaw_Kt3ZLTto-WtxWPFKpQrqx-H7KpgUyvxnWEQ&e=

I think it's still necessary in general - ENOTTY is what kernel vfio returns so
we need to align for that for any other use cases.

> > +
> > +# device.c
> > +vfio_user_device_io_device_feature(uint32_t argsz, uint32_t flags);
> 
> missing format here ^

Oops, surprised I didn't get errors with this

regards
john